Binary-decimal and decimal-binary routines for IEEE doubles